# What is Denoising autoencoders? Definition and meaning

Trains a network to restore clean data from a corrupted copy, which yields robust representations.

## Definition

**Denoising autoencoders**: Trains a network to restore clean data from a corrupted copy, which yields robust representations.

In short: Learning by reconstructing corrupted inputs.

## When it is used

Self-supervised learning and imputation on tables.

## Learn more

The full guide to **[Denoising autoencoders](/tools/denoising-autoencoders)** explains how it works, when to use it and how it compares with the alternatives, with examples. The [glossary](/glossary) lists every term on the site from A to Z.

Previous term: [Delta Lake](/glossary/delta-lake) · Next term: [DETR](/glossary/detr)
